Pyramids Empirical Evidence on the Costs and Benefits

Pyramids: Empirical Evidence on the Costs and Benefits of Family Business Groups Ronald W. Masulis, Peter K. Pham & Jason Zein

Introduction n Ownership & control of publicly listed firms by wealthy families and individuals is common in emerging markets n Wealthy families often control multiple independent firms through business group structures – Holding companies – Pyramids – Cross-shareholdings – Dual class shares n Several studies point to important costs and benefits of family business groups (Khanna and Yafeh, 2007 Survey)

Introduction Ultimate Investor Direct Control Middle layer Bottom layer Firm B Firm E Firm C Firm D Firm G Firm F

Costs and Benefits of Groups n Costs – Extraction of Private Benefits of Control (tunneling) • Bertrand, Mehta and Mullainathan (2002), Bae, Kang and Kim (2002), La Porta et al (2002), Claessens et al (2002), Joh (2003), Baek, Kang and Lee (2006) – Political influence and lobbying against legal and capital market reforms • Fogel (2006), Morck, Strangeland Yeung (2000), Morck Wolfenzon and Yeung (2005), Stulz (2005). – Social welfare costs of inefficient internal capital allocations • Almeida and Wolfenzon (2006 a)

Costs and Benefits of Groups n Benefits – Group reputation can substitute for weak legal and regulatory mechanisms • Fisman and Khanna (2004), Khanna and Rivkin (2001), Khanna and Palepu (1997, 2000) – Source of valuable equity capital – internal capital market. – Investment by minority shareholders • Almeida and Wolfenzon (2006 a, 2006 b) – Risk sharing mechanism, credit guarantees • Gopalan, Nanda and Seru (2007), Khanna and Yafeh (2005)

Motivation & Contribution n Family business groups a dominant fixture in emerging markets as well as developed markets. – Implication: Minority shareholders continue to co-invest with families within group structures, despite the risks. Why? – Are family-group firms good investment opportunities? n Examine relative costs and benefits of family business groups at: – Firm level: • Distinguish between family group firms and non-group firms • Examine business group structures and within-group positioning of firms. – Country Level: • Large number of sample countries and more extensive firm coverage compared with La Porta (1999), Claessens, Djankov, Fan and Lang (2002), Khanna and Yafeh (2005)

Ownership Data and Group Construction n Ownership data obtained for 27, 987 firms from 45 countries as of 2002 – Ownership databases: Osiris, Worldscope (others from Lexis. Nexis). Annual Reports: Mergent Online, stock exchanges and company websites n 1 st Step: Distinguish between widely held, directly controlled and pyramid controlled firms: – Closely held firm threshold is 20%, but it is dropped to 10% if the largest shareholder is a founder, CEO or has board control n 2 nd Step: Establish the identity of the controlling shareholders through media articles, stock exchanges, company websites, and annual reports

Ownership Data and Group Construction n 3 rd Step: Distinguish family group firms from independent family firms – Two or more firms in the same national market controlled by a common family are defined as belonging to a family business group n 4 th Step: Construct and verify the structure of each family group – Identify cross holdings within group – Identify the number of pyramidal layers and the position of each member firm – Calculate the cash-flow and control rights of each firm in the business group and adjusted for dual class shares n Procedure yields 845 family-controlled groups comprising 2526 firms.

Descriptive Statistics: Highlights Country N Sri Lanka 117 % % Group Pyramid % Group MC 63. 25 41. 68 48. 72 % % Market Pyramid Largest Group Cap MC 18. 19 SE Captain % Listed firms % 45. 25 Philippines 221 Listed firms 17. 19 to% Market cap. belonging a 30. 12 % 8. 98 belonging to a due family group & to family business held through controlled group firms Israel 226 39. 86 26. 11 54. 96 groups held 43. 96 in pyramid 10. 85 de Ayala 11. 73 Dankner 7. 13 Turkey 250 39. 6 18. 80 42. 04 16. 90 Sabanci Family 11. 72 Colombia 56 39. 29 28. 57 44. 27 21. 10 GEA Group 24. 50

Firm-Level Analysis Widely Held / Controlled alone (Industry Peers) Ultimate Investor Top Layer Firm B Firm C Firm H Middle layer Bottom layer Firm E Firm D Firm G Firm F

Firm-Level Univariate Results

Multivariate Results – Group Vs Non-Group n Multivariate OLS regressions for Tobin’s Q on group affiliation. – Controlling for Size, Beta, Leverage, Asset Growth, CAPEX, Dividend Yield, Asset Tangibility, Age, Analyst Coverage, Country Effects and Industry Effects n We address endogeneity of group affiliation choice using – Two stage IV regression and treatment effects model • Idiosyncratic risk as instruments for group affiliation, following Villalonga and Amit (2006). n Control for the benefits/costs of family and other control types – Repeat above regression using controlled firms only – Repeat above regression using family-controlled firms only

Multivariate Results – Group Vs Non-Group Analysis

Multivariate Results – Group Vs Non-Group (Endogeneity Correction)

Multivariate Results – Within-Group Analysis n Estimate Q regression on various group ownership and control measures. – Ultimate cash flow rights, control rights, group layer position and group direct ownership – Include (a) group-fixed effects to control for reputation and differences in entrepreneurial skill, and (b) industry fixed effects. With respect to Firm D: Family A 40% • Direct Ownership: 20%+40% = 60% 30% • Cash Flow Rights of A: 0. 4*0. 2 + 0. 3*0. 4 = 20% Firm C Firm B 50% Firm E 20% Firm D 40% 60% Firm F • Weakest Control Link: 20%+30% = 50% • Layer position: 1 (for D, E and F), 0 (for B and C)

Multivariate Results – Within-Group Analysis

Multivariate Results – Within-Group Analysis n Avoid systematic differences b/w group vs. non-group firms using Heckman correction for selection bias n Endogeneity of Q of ownership and control measures – IV regressions using idiosyncratic risk and same-industry indicator to instrument for individual ownership measures.

Country-Level Analysis n We measure several quantifiable country-level characteristics that could explain cross-country variation in family groups. – Private Benefits of Control / Corporate Governance • Corporate Governance Index, Block Premium, Competition, Tax Transparency, Newspaper Circulation (Newspaper) – Access to Capital • Savings to GDP, Institutional Funds, Log GDP per Capita – Taxation & Regulatory Factors • Consolidated Accounting of Group • Tax Transparency • Partial Acquisition Rules (Takeover Constraint Index) • Political Stability

Country-Level Regressions

Country-Level Robustness Checks n Controlling for unobserved determinants of the scope for private benefits of control: Repeat the country level regressions using – Group Firms / Controlled firms & Group Firms / Family Controlled Firms as the dependent variable – Block premium as an additional explanatory variable n Addressing reversal causality interpretations using instrumental variable (IV) regressions: – Legal Origin Corporate Governance Standards – Savings per GDP Institutional Funds n Combining country- and firm-level variables n Other potential determinants: alternative investor protection indices, professional manager availability, inheritance law, ownership disclosure rule n Results are qualitatively unchanged.

Conclusion n Country-Level – Access to capital, as well as regulation and taxation appear to be important in explaining business groups n Firm-Level – Lower Q for group firms may simply reflect endogenous selection, consistent with internal capital markets benefits of groups. – Correcting for endogeneity biases, group affiliation leads to an increase in value. – Placing firms at the bottom of pyramids increases their value due to access to internal and external capital and reputation benefits. – The direct ownership of the group appears to be the mechanism through which groups signal commitment and increase value rather than through their ultimate cash flow rights at the Apex.